This company has figured out how to make cheaper SWIR cameras

[frogg]

I think I get your interest in SWIR more clearly now. It's more of a "We need capability to image in every possible wavelength" point of view.

Just the simple idea that an object appears differently when viewing different spectra is good enough to warrant interest

[CatalinaWOW]

There are a myriad of reasons for selecting spectral bands.  Availability of lens and window materials that are compatible with the operating environment.  The size aperture required for diffraction limited performance.  The need for additional systems such a cooling or thermal stabilization.  Atmospheric transmission, which may favor absorbtion, or uniformity over time, or lack of absorbtion.  The list goes on and on.

In any field, from smart phones, to radios to military gear there is advantage to be gained by clever use of technical capabilities.  The winners are those that go where others haven't and find useful capabilities.

[PedroDaGr8]

Has anyone found patent applications or publications from this company which explain further details?

I am curious how the quantum dots actually work -- do they convert incoming light into visible-range photons or directly into electrons? How close is the CMOS architecture to a typical CMOS camera sensor? (Or is it indeed a standard camera chip, with just the quantum dot conversion layer added?)

The quantum efficiency and the noise level of their current products are somewhat disappointing. Regarding noise, I am surprised by the very high readout (?) noise baseline, which is there even at minimal exposure times. Yes, their chips are optimized for fast readout and are apparently uncooled, but why is the noise level that high?

I strongly doubt they are doing IR to visible light conversion, as that is stunningly inefficient. They would need a charge transfer/emissive layer which could absorb the excitons and allow for double exciton recombination (to provide the needed energies to go upwards in energy). This has been done using materials like ruberene and PbS quantum dots, but the QY was essentially 1.2%!

More likely, they are doing a direct charge transfer from the quantum dots to some sort of silicon substrate (maybe CMOS). In this motif, it would function as an array of tiny IR solar cells, providing an effective quantum yield up to the 20% range (not far off from the stated numbers here). As for the noise levels, heat is going to play a big role in the noise levels, as will design and structure. If they are transmitting the energy directly, this can induce more noise. Additionally, things like exciton relaxation time (i.e. does it recombine and emit light to adjacent pixels), isolation between the wells, etc.

Found this paper and it goes pretty deeply into the various types of quantum IR detectors: https://spie.org/samples/PM280.pdf

[CatalinaWOW]

There is lots of stuff out there, most behind a paywall. 

Here are a couple of free ones that nibble around the edges.

https://www.laserfocusworld.com/articles/print/volume-43/issue-12/features/quantum-dot-detectors-quantum-dot-ir-photodetectors-get-lsquohotterrsquo.html

https://apps.dtic.mil/dtic/tr/fulltext/u2/a596455.pdf

[Fraser]

Paywalls are so annoying ..... thank goodness for Sci-Hub 😄

[Max Planck]

Based on what was published so far, the company is using a standard, thin-film technology (thus lower costs), depositing PbS photodiodes directly on standard CMOS ROICs.

Max

[ebastler]

Based on what was published so far, the company is using a standard, thin-film technology (thus lower costs), depositing PbS photodiodes directly on standard CMOS ROICs.

Max

Where did you find that information? I did not see anything about that on their website.
Thanks!

[Max Planck]

For example here:
https://www.imveurope.com/analysis-opinion/colloidal-quantum-dot-sensor-disrupt-swir-imaging-market

https://www.imveurope.com/analysis-opinion/innovative-imaging-tech-shortlisted-vision-stuttgart-award

Max

[Marco]

I strongly doubt they are doing IR to visible light conversion, as that is stunningly inefficient.

I recently saw a paper claiming a breakthrough with an in cavity non-linear crystal, 20% QE for MWIR to 800 nm ... not visible, but close enough.

[ebastler]

Thank you!

So the fill factor, and hence the quantum efficiency, won't be getting great any time soon... But it is reassuring that these cameras are based on entirely standard CMOS sensors. So they could follow up with a TEC-cooled, slower readout, lower noise version pretty easily.

[CatalinaWOW]

All of this has potential, but until someone makes a bunch it is dream stuff.  HgCdTe was going to solve all detector problems until the reality of producing large arrays with acceptable defect rates and at acceptable costs crept in.  Same for several other technologies.  It isn't necessarily that the technology is bad either.  It is partly that we have been spoiled by the hundreds of billions of dollars spent on silicon fabrication technology and often blissfully unaware of the many, many problems that have been solved along the road.  And only a portion of those solutions apply to these detectors.  They will never have enough production volume to warrant the investment that has solved so many problems in silicon.

[PedroDaGr8]

I strongly doubt they are doing IR to visible light conversion, as that is stunningly inefficient.

I recently saw a paper claiming a breakthrough with an in cavity non-linear crystal, 20% QE for MWIR to 800 nm ... not visible, but close enough.

That structure is outside my realm of expertise as a former nanomaterials chemist. I was strictly speaking about Quantum Dot based non-linear conversion, which is asofar highly difficult and inefficient.

All of this has potential, but until someone makes a bunch it is dream stuff.  HgCdTe was going to solve all detector problems until the reality of producing large arrays with acceptable defect rates and at acceptable costs crept in.  Same for several other technologies.  It isn't necessarily that the technology is bad either.  It is partly that we have been spoiled by the hundreds of billions of dollars spent on silicon fabrication technology and often blissfully unaware of the many, many problems that have been solved along the road.  And only a portion of those solutions apply to these detectors.  They will never have enough production volume to warrant the investment that has solved so many problems in silicon.

That is ONE area where colloidal quantum dots excel, they can utilize a number of technologies currently employed in OLED manufacture (spin coat, die stamping, etc.) for manufacturing large arrays. That being said, as I mentioned earlier, they introduce their own suite of issues (the largest being how to get the charge out of the quantum dot, when most are designed to contain it in the center).

[Ben321]

This company has figured out how to make cheaper SWIR cameras ... my skepticist strongly doubts it will be any cheaper for the end customer.

You think it will just be more profit for the company that makes it, but same price to the end user?

[Ben321]

They will never have enough production volume to warrant the investment that has solved so many problems in silicon.

Actually "never" is too strong of a word. I see a future where SWIR, LWIR, and MWIR "point and shoot" cameras are as easy to find (and as cheap to buy) as visible light "point and shoot" cameras are now. Ever heard just some random person (not a professional artist) say "that's a beautiful sunset"? So art is a huge part of even amateur photography. There are 2 types of art, beauty based, and surrealism based. And what better way to make art of that second type (surrealistic art) than to take a photo of your everyday surroundings, but using an exotic wavelength. SWIR, MWIR, and LWIR are all capable of this. And I think that the average person is more interested (at least to some degree) in artistic photos, than most of these companies realize.

This isn't something that major IR camera producers recognize at this point, unfortunately. If they did, they would be investing in a way to mass produce SWIR, MWIR, and LWIR cameras of the cheap "point and shoot" variety, and sell them at every Costco, Target, and BestBuy store in the country.

As with other consumer level cameras that are not intended for scientific use, there is a much larger level of acceptable defects in the image sensor. Fewer sensors being rejected based on defects, would mean more cameras being sold, so each camera could be cheaper. Because the final audience is going to be large, the devices will need to be mass produced. This will be what motivates them to finally invest heavily in finding more efficient ways of manufacturing sensors of this type on a large scale.

I can easily see a future 10 to 20 years from now, where it will be possible to get a point and shoot camera for SWIR at 4K resolution, MWIR at 1080p resolution, or an LWIR camera at 720p resolution; each at the price that a normal visible light point and shoot camera is now. And the version of the camera sold in the US will actually have 30FPS frame rate (overseas versions of the cameras will get 9FPS instead).

[CatalinaWOW]

Never is probably too strong a word.  But the barriers are very real and the market not obvious.  NIR film was available for conventional cameras for decades and never developed much of a market, in spite of beautiful art shots.  I suspect game and surveillance camera will be the real market, but unfortunately with market sizes comparable to the image intensifier versions.  I hope and wish I am wrong, but I just don't see this happening soon.

[bugi]

As with other consumer level cameras that are not intended for scientific use, there is a much larger level of acceptable defects in the image sensor.

Actually, depending on the purpose, scientific use can actually accept much more poorer quality than consumers. Science doesn't care about a bunch of dead pixels at one spot, they'll just shift it a bit to use the area that works, and are happy to have saved half the camera price for exactly the same scientific results. The research project budgets are typically 50% too low to start with anyway.

There are certainly use cases where such flaws are showstoppers, too, and there are projects that have gotten all the funding they need and then some. The latter will be buying flawless unnecessarily good instruments just because they can. And bunch of art to the hallways and design chairs from Italy, etc. (Yeah, talking from experience, after watching neighbor lab burn more money on "waste" than our lab's whole budget.  You know, if they didn't waste all that money, they would have had extra left at the end of the year, and next year their budget would have been adjusted down, and the extra spread to other labs needing it more... "university politics & bureaucrazy" (misspelling intended).) 

[Ben321]

As with other consumer level cameras that are not intended for scientific use, there is a much larger level of acceptable defects in the image sensor.

Actually, depending on the purpose, scientific use can actually accept much more poorer quality than consumers. Science doesn't care about a bunch of dead pixels at one spot, they'll just shift it a bit to use the area that works, and are happy to have saved half the camera price for exactly the same scientific results. The research project budgets are typically 50% too low to start with anyway.

Usually though scientific cameras are built to a higher (not lower) standard by companies like FLIR. And thus the increased cost above equivalent consumer-level cameras. For example, not only is the total number of bad pixels kept below a certain threshold, but any VOx microbolometer that had 2 or more adjacent bad pixels is always rejected. It's tough-to-meet specs like that which mean 9 out of every 10 VOx microbolometers gets thrown in the trash, HUGELY increasing the cost of thermal imagers. Consumer level visible-light digital cameras are cheaper, because at worst they only throw out (at most) 50% of their CMOS or CCD chips, because the specs for consumer level cameras are much lower. The actual per-chip manufacturing cost (even with exotic materials like VOx) is not that high, but the price is raised dramatically because of how many chips intended for scientific cameras get tossed down the garbage chute for not meeting scientific-level specs.

[bugi]

As with other consumer level cameras that are not intended for scientific use, there is a much larger level of acceptable defects in the image sensor.

Actually, depending on the purpose, scientific use can actually accept much more poorer quality than consumers. Science doesn't care about a bunch of dead pixels at one spot, they'll just shift it a bit to use the area that works, and are happy to have saved half the camera price for exactly the same scientific results. The research project budgets are typically 50% too low to start with anyway.

Usually though scientific cameras are built to a higher (not lower) standard by companies like FLIR. And thus the increased cost above equivalent consumer-level cameras. For example, not only is the total number of bad pixels kept below a certain threshold, but any VOx microbolometer that had 2 or more adjacent bad pixels is always rejected. It's tough-to-meet specs like that which mean 9 out of every 10 VOx microbolometers gets thrown in the trash, HUGELY increasing the cost of thermal imagers. Consumer level visible-light digital cameras are cheaper, because at worst they only throw out (at most) 50% of their CMOS or CCD chips, because the specs for consumer level cameras are much lower. The actual per-chip manufacturing cost (even with exotic materials like VOx) is not that high, but the price is raised dramatically because of how many chips intended for scientific cameras get tossed down the garbage chute for not meeting scientific-level specs.

Yeah, quality costs - no matter who ends up needing it. I was mostly pointing out that science vs. consumer doesn't directly dictate better vs. lower quality in all cases. Though companies may indeed use such differentiation for marketing reasons.